The New Studios Aren't Studios

In April 2023, the Republican National Committee released an entirely AI-generated attack ad within hours of Joe Biden's reelection announcement. No camera crew. No location scout. No editor's note in the credits. A dystopian campaign spot built end-to-end by a model and a prompt, distributed before the news cycle on the actual announcement was done.

That ad is a useful anchor -- not because it was good, but because it was first. It was the moment the production cost of a political TV spot effectively went to zero. Every category downstream of that -- film, advertising, music, news, content -- is now on the same curve.

What followed in the next 24 months reads like a category being beta-tested in public:

• The DeSantis campaign ran AI-generated images of Trump hugging Anthony Fauci in a June 2023 attack ad.
• A political consultant was criminally charged and fined by the FCC for AI-cloned Biden robocalls telling New Hampshire voters not to show up to the primary.
• Donald Trump posted AI-generated images of a fake Taylor Swift endorsement in August 2024.
• By the end of 2024, NPR's wrap-up of the election cycle documented AI-generated political memes and synthetic videos as a permanent feature of campaigns across dozens of countries.

And outside campaign infrastructure, individual operators figured out the same architecture independently. Spencer Pratt -- yes, that Spencer Pratt -- built a one-person political content operation out of his phone after the January 2025 Palisades fire, racked up tens of millions of views, took a public-facing battle with Governor Gavin Newsom directly to DC, and ended up meeting with Attorney General Pam Bondi and other federal officials. Whatever you think of him, the architecture is the point: one operator, three subscriptions, a phone, and a real political outcome.

For eighty years, creative production was structured around scarcity. Cameras were expensive. Talent was rare. Distribution was gatekept. Generative AI has collapsed all three. The unit of "studio" is no longer a building. It's a stack of three subscriptions and one operator.

The next Pixar, the next Wieden+Kennedy, and the next Lincoln Project may all be the same kind of company. They just point at different audiences.

Three categories converging into one

The companies don't look alike on the surface. Under the hood, they're running the same playbook.

The film and series studios

• Promise -- gen AI film and series studio. Backed by a16z and Peter Chernin's North Road, with Google's AI Futures Fund and Michael Ovitz's Crossbeam as strategics. Founded by George Strompolos (ex-CEO of Fullscreen), Dave Clark, and Jamie Byrne.
• Asteria Film -- premium nonfiction studio with an applied-AI research division. Owned by XTR.
• Toonstar -- AI-assisted animation studio behind YouTube's StEvEn & Parker. Signed with WME.
• MOTHER.Tech -- Brooklyn-based, raised ~$15M to launch Degen, a prompt-less AI content app.

The app studios

• Athenix, Pory, Agentic Studio, Virix Labs, Keldrix -- venture studios shipping multiple AI-native consumer and prosumer apps in parallel.

The thesis is portfolio-style: build 10 small focused apps, let SEO/ASO sort them, double down on the winners. The cost per shot has dropped enough that you can take more shots than the previous generation of venture studios ever could.

The political and influence shops

This is where the curve is sharpest and the regulation is loosest.

• CampaignAI.us -- AI video and messaging for down-ballot campaigns, priced below traditional agency fees.
• Rapid Signal -- verified AI-generated content for campaigns, 4-8 video clips per day.
• PoliticalTesting.com -- message testing against synthetic voter audiences.
• Electric Twin -- £10M raised, "synthetic audiences" of human thought, founded by former Downing Street advisors.
• And the long tail of operators -- individuals running AI-assisted content shops out of laptops, monetizing through PACs, TikTok payouts, and brand sponsorships. Spencer Pratt's Palisades fire accountability campaign is the most legible recent example -- a reality-TV personality who, in less than a year, built a political content operation visible enough to land meetings with the U.S. Attorney General. He's not the story. The story is that the toolchain works for anyone willing to ship daily.

What they all share

All three groups are doing the same five things:

1. Replacing expensive production -- talent, cameras, editors -- with a model plus a prompt.
2. Compressing creative iteration from weeks to hours.
3. Closing the loop between generate → test → optimize → distribute in a single workflow.
4. Building "studios" with 3-10 people instead of 300.
5. Treating their output as performance content, measured against engagement, not Cannes Lions.

Squint at Promise and squint at CampaignAI.us. The architecture is identical. One sells to streamers; the other sells to PACs. The cost structure is the same.

The political ad story is the leading indicator

Political content has the tightest feedback loop in media -- there's a literal election day -- and the loosest regulatory enforcement. The FEC is years behind on AI disclosure rules. That combination makes political the canary for everything else.

Whatever's working in a 2026 Senate race shows up in your brand campaign in 2027 and your local news feed in 2028.

The investable layers

• Picks and shovels. Testing, compliance, and verification -- who watermarks, fact-checks, and disclaims this stuff? TruePic and C2PA-aligned provenance tools are early and likely to be acquired by every platform that needs an answer for regulators.
• Specialized studio software. Vertical tools for specific creative motions -- animation, music with rights cleared, ecommerce video, political-ad compliance.
• Studios with real distribution. The rare combination of generative production plus a built-in audience or buyer (Promise + Chernin, Asteria + XTR).
• Defense. Detection, deepfake forensics, election integrity. As critical as offense. The companies that build content provenance infrastructure for the next election cycle will be the cybersecurity primes of the 2030s.

Three open questions

• Is the moat the model, the talent, the distribution, or the brand? Almost certainly the last two. But the first companies to win the first three get the cleanest shot at brand.
• What is "consent" when any voice can be cloned and any face can be regenerated? Tennessee passed the ELVIS Act. The EU is moving. The companies that build verifiable consent layers win a category nobody's named yet.
• Does the app-studio model produce venture-scale outcomes, or a lifestyle business in disguise? Ten apps doing $500K ARR each is a great living, not a venture return. The selection function is still unknown.

Kicker

Four articles. Four categories. Kid Tech. Elder Tech. New Sports. Creative AI Studios.

All four are being rebuilt by the same architectural shift -- AI as the personality, hardware or software as the wrapper, trust and distribution as the moat, an old-economy buyer (parents, adult children, recruiters, PACs) as the customer.

The hard part of investing in 2026 isn't picking which of these matters most. It's realizing they're all the same trade.